Latent State Estimation for Financial Time Series, Estimating Financial Health with MCMC Methods and Particle Filters

Abstract

Financial Modelling allows for prudent decision making for individual business owners and other stakeholders. The Financial Health can be seen as an underlying measure which governs the companies ability to meet its obligations and make profits. Therefore Financial Health is linked to the company’s cash flow which can readily be observed. We consider the Financial Health as a dynamic latent state and infer it from the cash flow. We are estimating this latent state under the Bayesian paradigm to take stylized properties of the cash flow into account, using a Particle Filter as part of a Monte Carlo method to sample the posterior distribution of latent state and model parameters. We investigate the performance of this approach on a real data set consisting of real cash flow from small Swedish businesses.